Recall of paired-associate lists (declarative memory) and mirror-tracing skills (procedural memory) was assessed after retention intervals defined over early and late nocturnal sleep. In addition, effects of sleep on recall were compared with those of early and late retention intervals filled with wakefulness. Twenty healthy men served as subjects. Saliva cortisol concentrations were determined before and after the retention intervals to determine pituitary-adrenal secretory activity. Sleep was determined somnopolygraphically. Sleep generally enhanced recall when compared with the effects of corresponding retention intervals of wakefulness. The benefit from sleep on recall depended on the phase of sleep and on the type of memory: Recall of paired-associate lists improved more during early sleep, and recall of mirror-tracing skills improved more during late sleep. The effects may reflect different influences of slow wave sleep (SWS) and rapid eye movement (REM) sleep since time in SWS was 5 times longer during the early than late sleep retention interval, and time in REM sleep was twice as long during late than early sleep (p < 0.005). Changes in cortisol concentrations, which independently of sleep and wakefulness were lower during early retention intervals than late ones, cannot account for the effects of sleep on memory. The experiments for the first time dissociate specific effects of early and late sleep on two principal types of memory, declarative and procedural, in humans.